Not everything manual is worth automating, and treating automation as a universal upgrade is how companies end up with a dozen brittle scripts nobody maintains.
The filter that actually works: is the task repetitive, rule-based, and high-volume? If yes to all three, automate it — the return is immediate and compounding. If it needs judgment on a case-by-case basis, automating it just moves the judgment call somewhere harder to inspect, not away.
Automate the boring, high-volume work first. Everything else is a decision, not a script.
